That has two issues:

1. The water they want to measure comes from a different source than the lab water. Subtracting the lab water's contamination from that of the measured water would make no sense, since you'd get different results for the measured water's content depending on the lab water's purity.
2. You need a zero baseline to calibrate your measurements. No matter which method you use to measure, you'll always have a measuring error. Which means your measurement will show some amount even with perfectly pure water.
   Calibration with pure water will tell you X result is the systematic measuring error, so that is what you do subtract.

I don't know the specifics of his experimental setup, only what he told me, but my guess is that the differences between the measured water and the control water weren't clear enough to get a statistically significant result.